What if your geolocated IP is from … a VPN? Maybe one outside the jurisdiction of the law?

Is it hard to beat on performance for the cost, though?

I’m no ML engineer and far from an LLM expert. Just reading the article though it seemed to me that leveraging an SQL database here was a bigger issue than using traditional ML on the data, rather than the LLM being a win specifically. Just finding anything that was better suited than string matching on a RDBMS to the type of inputs seems like the natural conclusion when the complaint in the article itself was literally about SQL.

To be a lumper for a second, all models are flawed. But some are useful.

I like your definitions, but all three of these could be called subsets of compilers.

By the definitions given they can not, as no function subsumes another. By whatever you define as "compiler" maybe, but I see no point in this kind of interaction that essentially boils down to subsumtion to an entity you refuse to describe any further.

Is there a merit to this? Can whatever you call compiler do more? Is it all three of the things mentioned combined? Who knows - as is stands I only know that you disagree with the definitions given/proposed.

I think they are fine definitions. I think a transpiler, a term rewriter, an assembler, a stand-alone optimizer, and even some pretty printers are subclasses of compilers.

I define a compiler as something that takes an input in a language, does transformations, and produces a transformed output in a language. All of them do that, and they are more specific terms for types of compilers.

Except that they do what useful words do; provide (more) useful information.

Fair. I don’t believe I said they were useless terms for differentiation of types of compilers, though. I just said they can all be thought of as a class as different types of compilers.

So what? A car, a bike, and a truck can all be called subsets of vehicles, but we still have (and need) different words for each type.

Primates, canids, felids, and ungulates are all subsets of mammals and all have further subsets. Mammalia is further a subset of animalia. When we’re discussing categorizations, it’s often helpful to have multiple levels of categories. I’m not sure why you seem to be calling out specificity as a gotcha, when my argument isn’t at all that we don’t need multiple terms. It’s that we should consider these things in terms of similarity and specific differences, not throw away a term as useless as the article and its headline suggest.

A few anti-cheat systems rather than inspecting the local machine look for things like impossibly fast target acquisition in FPS games, or the server noticing when a shot is taken on an opponent who’s supposed to be totally obscured. Those aren’t perfect, but they don’t require kernel-level anticheat.

Cheating detection server side is expensive and probabilistic at best, kernel level anti cheat is a purely financial decision

I can’t speak for brendo, but I do most of my gaming on a separate PC-class machine from my home workstation, both of which are separate from my work laptop and personal laptop.

Honestly, as much as I prefer Linux for most things Linux staying pat and macOS stealing market share from Windows is almost as good as Linux taking those users. I think we’re currently seeing a trend starting of people leaving Windows for each of them, and some users for both of them.

The x86 running Windows isn’t perfect. The x86 rack system running Linux isn’t perfect. Android isn’t perfect. The Ford F150 isn’t the perfect pickup. Budweiser is far from the perfect beer.

The phrase “worse is better” has a lot of historical significance in computing. Long before that, though, Adolphus Busch started his brewing empire. If you take a brewery tour at an Anheuser-Busch brewery, they’ll tell you that the company’s flagship product, the aforementioned Budweiser, was never intended to be anyone’s favorite beer.

That’s right. One of the top selling beers in the world was never intended to be a personal favorite of a single buyer or beer drinker. What it was designed to be was unobjectionable, approachable, and good enough to serve your guests when their preferred beer runs out. There are so many varieties of beer that are so different, and they are often loved by some and despised by others. So an intentionally unremarkable but quality beverage was marketed to be a very popular second or third choice.

If most households have a Playstation and a Deck or Frame, or have a Switch and a Frame, or have a PC and a Deck then in total numbers the Steam machines just might be the top seller even if it’s not a universal favorite.

That sounds like a feature request to Google, not an indictment of the NYT.

Same way that a feature request to the police is not an indictment of the criminals.